---
title: "The Demographic Erosion of Fiscal Leverage: Twin Deficits in an Aging World"
author: "Brian Peters"
date: "February 2026"
version: "20260224_r2"
abstract: |
  The twin deficits hypothesis treats the fiscal-current account coefficient as stable across countries. We show it varies by a factor of twelve with demographic structure: from 0.71 in young economies to 0.061 in aged ones. This erosion operates through a channel mismatch. The demographic Z vector transmits entirely through the non-trade component of the current account (Z1 = 41.4, p = 0.007), while fiscal policy transmits through the trade balance. Fiscal leverage is empirically stronger when trade dominates the CA (fiscal x trade-dominance = 0.262, p < 0.001), and the absolute amplitude of trade-channel variation collapses in aging economies. Demographics do not confound the twin-deficit relationship (0.8% attenuation); they moderate it. Financial openness further erodes---rather than restores---fiscal leverage in aging economies (fiscal x KAOPEN = -0.068, p = 0.003 within the late-transition tercile). These findings imply that the IMF's External Balance Assessment framework should condition its fiscal coefficient on demographic transition stage: a uniform coefficient systematically overstates fiscal leverage in aging economies and understates it in young ones. Post-dividend economies face a fiscal leverage trap in which the need for external adjustment rises precisely as the primary policy tool loses effectiveness.
keywords: "twin deficits, demographics, current account, fiscal policy, external balance assessment, aging"
jel: "F32, F41, H62, J11"
---

# 1. Introduction

The "twin deficits" hypothesis---that fiscal deficits cause current account deficits---is one of the most influential propositions in open-economy macroeconomics. The logic is transparent: from the national accounting identity $CA = (S - I) - (G - T)$, a fiscal deficit ($G > T$) worsens the current account, all else equal, unless offset by a rise in private savings or a fall in investment. This transmission channel is central to the IMF's External Balance Assessment (EBA) methodology, which treats the fiscal coefficient as a stable parameter across countries and over time (Phillips et al., 2013; IMF, 2019).

We show that this uniformity assumption is dramatically wrong. Using an expanded panel of up to 170 countries over 1990--2024, we demonstrate that the fiscal--current account coefficient varies by a factor of twelve depending on demographic structure: from 0.71 in demographically young economies to a barely significant 0.061 in aged ones. In other words, a one-percentage-point fiscal consolidation in a young economy improves the current account by 0.71 percentage points of GDP, but the same consolidation in an aging economy delivers essentially nothing.

The mechanism we identify operates through the composition of the current account itself. A preliminary decomposition reveals that the demographic Z vector---an orthogonal polynomial representation of the full age distribution following Higgins (1998)---transmits entirely through the non-trade component of the current account ($Z_1 = 41.4$, $p = 0.007$), with a complete null on the trade balance ($Z_1 = -3.0$, $p = 0.878$). Demographics operate through the non-trade component of the current account (dominated by but not identical to investment income and transfers), not goods trade. Fiscal policy, by contrast, operates primarily through aggregate demand and the trade balance. Fiscal leverage is empirically strongest when trade dominates the CA (fiscal$\times$trade-dominance = $0.262$, $p < 0.001$), and the absolute amplitude of external balance variation collapses in aging economies (|Trade/GDP| falls from 15 to 7 across demographic terciles). This channel mismatch---fiscal policy targeting a lower-amplitude trade channel in aging economies where non-trade dynamics increasingly matter---explains why fiscal leverage erodes with demographic maturity.

Despite this erosion, we find that demographics and fiscal balance are nearly orthogonal as predictors of the current account: controlling for the Z vector attenuates the fiscal coefficient by only 0.8%. Demographics do not confound the twin-deficit relationship---they moderate it. The distinction matters for policy. Post-dividend economies face what we term a "fiscal leverage trap": the demographic tailwind has reversed (lifecycle dissaving now worsens the external position), fiscal policy has lost its effectiveness over external balances, and yet fiscal discipline remains the most direct policy lever for external balance management, even as its effectiveness diminishes. Each unit of consolidation yields a smaller external adjustment precisely when the need for adjustment is greatest.

Our findings have direct implications for the EBA framework and for sovereign debt sustainability assessments. External adjustment recommendations that assume a universal fiscal coefficient will systematically overestimate the effect of consolidation in aging economies and underestimate it in young ones. We provide evidence that the coefficient should be conditioned on demographic transition stage. Financial openness does not restore fiscal leverage in aging economies; if anything, it further loosens the fiscal--CA link ($p = 0.003$ within the late-transition tercile).

The paper proceeds as follows. Section 2 reviews the twin deficits and CA decomposition literatures. Section 3 describes the data and methodology. Section 4 presents the CA decomposition that motivates the fiscal analysis. Section 5 contains the twin deficits results. Section 6 discusses policy implications, and Section 7 concludes.

# 2. Literature Review

## 2.1 The Twin Deficits Hypothesis

The twin deficits proposition dates to the Reagan-era US experience, where large fiscal deficits coincided with large current account deficits. The theoretical foundation rests on the Mundell-Fleming model and its intertemporal extensions: a fiscal expansion raises domestic absorption, appreciates the real exchange rate, and worsens the trade balance. Early empirical work by Bernheim (1988) found support in a panel of OECD countries, while subsequent studies have debated the magnitude and robustness of the relationship.

Corsetti and M\"uller (2006) emphasize that the fiscal--CA link depends on the persistence of fiscal shocks and the degree of international risk sharing. Abbas et al. (2011) find a fiscal coefficient of approximately 0.3--0.5 in a broad cross-country panel, though with substantial heterogeneity. Chinn and Prasad (2003) show that the fiscal balance is one of the most robust determinants of the current account in medium-run regressions. The IMF's EBA methodology (Phillips et al., 2013) codifies this by including the cyclically-adjusted fiscal balance as a standard regressor in its CA norm equations, treating the coefficient as common across countries.

A key limitation of this literature is the assumption of parameter homogeneity. While some studies allow for income-group differences, no prior work systematically conditions the fiscal coefficient on demographic structure. This is the gap our paper fills.

## 2.2 Demographics, Fiscal Pressure, and External Balances

The lifecycle hypothesis predicts that population aging affects both fiscal balances (through pension and health spending) and current accounts (through aggregate savings). Auerbach and Kotlikoff (1987) pioneered generational accounting frameworks showing how demographic transitions create fiscal pressures. Lee and Mason (2011) document national transfer accounts across countries, revealing how age-specific consumption and production patterns generate fiscal obligations.

Our companion paper on fiscal dominance (Peters, 2026e) shows that aging threatens fiscal sustainability through the spending channel, not interest rates: a 10-percentage-point increase in the old-age dependency ratio is associated with 12 percentage points higher expenditure but only 5 percentage points higher revenue. The question this paper addresses is whether demographics *confound* the fiscal--CA relationship (acting as a common cause of both) or *moderate* it (altering the transmission mechanism).

## 2.3 Current Account Decomposition

The current account identity ($CA = TB + NTR$, where $TB$ is the trade balance and $NTR$ is the non-trade residual comprising income and transfers) provides a natural decomposition. Obstfeld and Rogoff (2005) argue that valuation effects and investment income have become increasingly important for understanding external positions. Gourinchas and Rey (2007, 2014) develop the "exorbitant privilege" framework showing how the composition of foreign assets and liabilities generates predictable income flows. Lane and Milesi-Ferretti (2007) document the secular rise in gross foreign asset positions, implying that net investment income has grown as a share of the current account.

Our contribution is to show that demographic effects operate through different CA components depending on the demographic variable used. The Z polynomial vector (capturing the full age-distribution trajectory) transmits through non-trade channels, while the OADR (capturing the contemporaneous elderly share) transmits through the trade balance. This decomposition motivates the fiscal analysis: if demographics work through non-trade channels but fiscal policy works through trade, the transmission mismatch explains why fiscal leverage erodes in aging economies.

## 2.4 EBA Treatment of the Fiscal Coefficient

The IMF's External Balance Assessment methodology, introduced in Phillips et al. (2013) and updated in IMF (2019), includes the cyclically-adjusted fiscal balance as a regressor in both current account and real effective exchange rate norm equations. The fiscal coefficient is estimated from the pooled panel and applied uniformly across countries to generate CA "norms" and assess over- or under-valuation. Our finding that this coefficient varies by a factor of twelve across demographic stages implies that EBA norms systematically misattribute demographic effects in the CA to fiscal policy gaps.

# 3. Data and Methodology

## 3.1 Panel Construction

We employ the expanded panel dataset from the multilateral followup study, comprising up to 170 countries over 1990--2024 (varying by specification and data availability) with current account balances (CA/GDP), the Chinn-Ito financial openness index (KAOPEN), NFA positions, and standard EBA controls. Our demographic variables are orthogonal polynomial projections $Z_1, Z_2, Z_3$ of the full age distribution, following Higgins (1998) and Koomen and Wicht (2023). Direct dependency ratios (old-age dependency, youth dependency) provide complementary specifications.

To decompose the current account, we supplement the panel with the World Bank's external balance on goods and services (NE.RSB.GNFS.ZS), which measures net exports of goods and services as a share of GDP. The *non-trade residual*---defined as $\text{CA/GDP} - \text{Trade Balance/GDP}$---captures everything in the current account not attributable to goods and services trade: net primary income (investment returns, compensation), net secondary income (remittances, transfers), capital transfers, and statistical discrepancy. We use "non-trade residual" throughout when referring to this constructed variable, reserving "income balance" for official BOP income data (primary + secondary).[^income-note]

[^income-note]: The non-trade residual correlates at $r = 0.950$ with official BOP income data (WDI BN.GSR.FCTY.CD + BN.TRF.CURR.CD, $N = 5{,}748$, 173 countries), confirming it is dominated by non-trade flows rather than measurement noise. However, Z regression coefficients differ between the two measures (see Section 4.2.1 and Table A1), so we do not claim the residual identifies a specific investment-return mechanism.

For the twin deficits analysis, we use the cyclically-adjusted fiscal balance as a share of GDP. Demographic tercile splits are constructed using the mean old-age dependency ratio over the sample period, dividing countries into early-transition (youngest third, 77 countries), mid-transition (78 countries), and late-transition (oldest third, 55 countries) groups. Tercile assignment is time-invariant (based on each country's sample-period mean) to avoid endogeneity from countries moving between groups as they age, and to prevent compositional changes from contaminating within-tercile estimates.

## 3.2 Estimation

All models are estimated by Panel GLS with Prais-Winsten AR(1) correction, country and year fixed effects, following the EBA methodology.

## 3.3 Estimands and Limitations

All specifications are descriptive panel associations with entity and time fixed effects plus AR(1) correction. We identify conditional correlations, not causal effects. The CA decomposition exploits an accounting identity; the twin deficits analysis conditions on observed fiscal balances without instrumenting. The companion causal identification paper (Peters, 2026c) addresses endogeneity directly; results here should be interpreted as establishing robust stylized facts and conditional associations rather than identifying structural parameters.

# 4. The Demographic Anatomy of the Current Account

This section establishes *how* demographics affect the current account---through which component---before turning to the fiscal interaction in Section 5. The decomposition is essential for understanding *why* fiscal leverage erodes: demographics and fiscal policy operate through different channels.

## 4.1 The Z Vector: Non-Trade CA Dominance

We estimate:

$$Y_{it} = \gamma_1 Z_{1,it} + \gamma_2 Z_{2,it} + \gamma_3 Z_{3,it} + \beta' X_{it} + u_{it}$$

where $Y$ is alternately CA/GDP, the trade balance, or the non-trade residual, and $X$ includes the standard EBA controls (fiscal balance, NFA lag, relative productivity, KAOPEN).

This paper does not seek to re-estimate the multilateral Z$\to$CA baseline; we use the decomposition to identify which CA component is most aligned with demographic structure in the richer control specification relevant for fiscal transmission.

The decomposition using the Higgins Z vector yields a striking result. In the baseline (M1), $Z_1$ on the overall current account is $\hat{\gamma}_1 = 3.0$ ($p = 0.805$), which is not significant in this specification---a result that reflects the inclusion of both relative productivity and KAOPEN as controls in the expanded 170-country sample, which absorb much of the demographic signal (the companion 140-country paper obtains $Z_1 = 38.4$, $p = 0.017$ in a parsimonious specification without log relative productivity). Decomposing this near-zero overall effect reveals that the signal operates entirely through the non-trade CA component: $Z_1$ on the non-trade component yields $\hat{\gamma}_1 = 41.4$ ($p = 0.007$), while $Z_1$ on the trade balance is essentially zero ($\hat{\gamma}_1 = -3.0$, $p = 0.878$). All three Z components ($Z_1$, $Z_2$, $Z_3$) are significant for the non-trade component (all $p < 0.01$) and completely null for the trade balance.

An important caveat applies to the non-trade residual. Our residual correlates at $r = 0.950$ with official BOP income data, confirming it is not noise. However, when we re-run the Z regressions using official BOP primary income directly, $Z_1$ enters with the *opposite* sign ($-31.1$, $p = 0.009$) and the overall $R^2$ is near zero. This divergence arises because the residual includes statistical discrepancy and capital account items that the official BOP income measure excludes, BOP primary income coverage is sparser (particularly for developing countries), and the accounting identity used to construct the residual mechanically links it to the CA. Table A1 in the appendix summarizes these differences.

We therefore frame the result as: *demographics operate through the non-trade component of the current account*, which is dominated by but not identical to net investment income and transfers. The trade balance channel is unambiguously null for Z, regardless of measurement approach.

## 4.2 The OADR: Trade Balance Dominance

A complementary specification using the old-age dependency ratio directly reverses the pattern. OADR predicts the trade balance strongly ($\hat{\beta} = -42.0$, $p < 0.001$) and the non-trade residual more weakly ($\hat{\beta} = 9.8$, $p = 0.074$). The trade effect is negative: higher OADR implies a worse trade balance, consistent with aging populations consuming more relative to production.

This apparent contradiction between Z and OADR specifications is informative. The Z vector captures the full age-distribution shape---including the lifecycle savings position---which operates through cumulative non-trade channels. The OADR captures only the elderly share, which directly affects consumption-to-production ratios (the trade channel). The distinction implies that the *level* of aging (high OADR) worsens the trade balance, while the *trajectory* of aging (captured by Z's polynomial representation of demographic transition) generates non-trade surpluses, plausibly through prior asset accumulation and income on foreign positions, though the specific mechanism within the non-trade residual cannot be isolated (see Section 4.1 caveat and Table A1).

## 4.3 KAOPEN Interactions and Channel Composition

Adding Z$\times$KAOPEN interactions reveals a conditional correlation consistent with a compositional channel: capital account openness is associated with a redistribution of demographic effects between CA components. The interaction $Z_1 \times KAOPEN$ is positive and significant for the trade balance ($\hat{\phi} = 21.5$, $p = 0.035$) and negative for the non-trade residual ($\hat{\phi} = -14.2$, $p = 0.098$). In financially open economies, demographic surpluses shift from the non-trade residual toward the trade balance.

This pattern is robust to controlling for exchange rate regime flexibility (all Z$\times$ERS interactions are null, $p > 0.69$), confirming that the compositional shift operates through the financial openness channel rather than nominal regime choice. However, when trade openness interactions are added, the Z$\times$KAOPEN coefficients on the trade balance lose significance ($p > 0.52$) while those on the non-trade residual remain significant ($p < 0.001$), suggesting that the non-trade channel reflects genuinely financial mechanisms.

## 4.4 NFA Creditor/Debtor Asymmetry

We allow asymmetric NFA effects by replacing $\text{NFA}_{t-1}$ with a spline split at zero: $\text{NFA}^+ = \max(\text{NFA}_{t-1}, 0)$ and $\text{NFA}^- = \min(\text{NFA}_{t-1}, 0)$. For the non-trade residual (M7), the creditor coefficient is $-1.09$ ($p < 0.001$) and the debtor coefficient is $-0.20$ ($p = 0.542$). The negative creditor coefficient means that *within* creditor countries, further NFA accumulation is associated with a *lower* non-trade residual---consistent with diminishing marginal returns on foreign assets, valuation adjustments, or mean reversion. For debtors, the near-zero coefficient ($-0.20$, $p = 0.542$) reflects offsetting forces: higher debt obligations generate income outflows, but debtor countries also tend to have higher-return asset portfolios and receive compositional shifts from FDI to debt as NFA declines, dampening the net relationship. Controlling for the world real bond yield in the creditor subsample attenuates the coefficient by only 12%, and interacting creditor NFA with the world rate is insignificant ($p = 0.80$), so the negative sign is not attributable to global yield compression.

Crucially, $Z_1$ remains significant at $39.2$ ($p = 0.011$) conditional on the NFA split---consistent with demographics capturing non-trade effects beyond what NFA/GDP measures. The specific mechanism (asset composition, return differentials, or transfer patterns) cannot be isolated given the sign reversal between our non-trade residual and official BOP primary income (Table A1).

## 4.5 Direct Evidence for Channel Mismatch

The decomposition in Sections 4.1--4.4 shows that demographics and fiscal policy operate through different CA components. We now test whether this channel mismatch directly predicts fiscal leverage, as the mechanism story requires. We define a *trade-dominance indicator* equal to one when the trade balance accounts for more than half of total CA component variation in a given country-year: $\mathbb{1}\{|TB_{it}/\text{GDP}| > |NTR_{it}/\text{GDP}|\}$, where $NTR$ is the non-trade residual. Interacting the fiscal balance with this indicator, the fiscal$\times$trade-dominance coefficient is $0.262$ ($p < 0.001$): the fiscal--CA link is stronger by 0.26 percentage points when trade dominates the current account. The continuous specification confirms this: fiscal$\times$trade-share yields $0.427$ ($p < 0.001$). In a subsample split, the fiscal coefficient is 0.67 in trade-dominant observations versus 0.41 in non-trade-dominant observations (Table R4).

This directly demonstrates the channel mismatch mechanism. Fiscal policy transmits to the current account through the trade channel; when that channel accounts for a larger share of external balance variation, fiscal leverage is correspondingly stronger.

## 4.6 Robustness: Winsorized Decomposition

Because trade balances exhibit extreme outliers, we verify the decomposition under symmetric winsorization. At p1/p99, the core pattern holds: $Z_1$ on the trade balance remains null ($5.1$, $p = 0.792$), $Z_1$ on the non-trade residual weakens to $25.7$ ($p = 0.078$), and OADR on the trade balance remains strongly significant ($-34.9$, $p < 0.001$). At the aggressive p5/p95 threshold, the $Z_1$ non-trade result loses significance ($11.4$, $p = 0.369$), while the trade balance null ($10.6$, $p = 0.534$) and OADR$\to$trade result ($-21.8$, $p = 0.001$) are preserved. We conclude that the Z$\to$trade null and OADR$\to$trade result are robust to outliers; the Z$\to$non-trade result is sensitive to influential observations, consistent with the role of extreme investment income observations in creditor economies (Table R1).

## 4.7 Implications for Fiscal Transmission

The decomposition and channel mismatch evidence establish the foundation for the fiscal analysis. Demographics operate through the non-trade CA component---dominated by but not identical to investment income and transfers---while fiscal policy operates primarily through aggregate demand and the trade balance. Fiscal leverage is directly stronger when trade dominates the CA (Section 4.5). While the *share* of CA accounted for by trade does not systematically decline with aging (Table R2--R3), the absolute amplitude of both components collapses (|Trade/GDP| falls from 15.1 to 7.1 across terciles), and aging economies spend more time in non-trade-dominant states where fiscal leverage is empirically lower. This channel mismatch---fiscal policy targeting a lower-amplitude margin in economies where non-fiscal dynamics increasingly matter---is the mechanism through which aging erodes fiscal leverage, as we demonstrate in Section 5.

# 5. Demographics and the Twin Deficits

## 5.1 Does Z Confound the Twin Deficit?

The baseline twin-deficit regression without demographic controls yields a fiscal coefficient of $\hat{\beta} = 0.526$ ($p < 0.001$, M1). Adding Z controls reduces this marginally to $0.522$ (M2). The attenuation is $(0.526 - 0.522)/0.526 = 0.8\%$---economically negligible. Demographics and fiscal balance are essentially orthogonal as predictors of the current account. The twin-deficit correlation is genuine, not an artifact of shared demographic origins.

This near-zero confounding is itself a finding. Despite the lifecycle hypothesis predicting that demographics should affect both fiscal balances (through pension/health spending) and current accounts (through aggregate savings), the two channels are nearly independent conditional on country and year fixed effects. The within-country variation in fiscal balances that drives the twin-deficit coefficient is largely orthogonal to the within-country demographic trajectory.

## 5.2 Aging Weakens Fiscal Transmission

The more consequential finding emerges from interactions. The fiscal$\times$old-age dependency interaction is strongly negative: $\hat{\delta} = -2.76$ ($p < 0.001$, M4). With the triple interaction including KAOPEN (M5), the fiscal$\times$OADR coefficient strengthens to $-3.46$ ($p < 0.001$), with a positive fiscal$\times$OADR$\times$KAOPEN term ($0.44$, $p = 0.056$) that initially suggested financial openness partially restores fiscal transmission. However, a more granular probe (Table R5) reveals that the triple interaction is misleading. In a 3$\times$3 factorial crossing demographic and KAOPEN terciles, the fiscal coefficient in aging economies is 0.060 (NS) for closed, 0.172 ($p = 0.009$) for mid-open, and $-0.018$ (NS) for fully open countries. Within the late-transition tercile, fiscal$\times$KAOPEN is $-0.068$ ($p = 0.003$)---openness *further erodes* rather than restores fiscal leverage. The marginal significance of the full-sample triple interaction likely reflects the strongly positive fiscal$\times$KAOPEN relationship in young economies ($0.085$, $p = 0.028$) leaking into the three-way term.

The interpretation is intuitive: in aging societies, fiscal consolidation generates smaller current account improvements because the aging population dissaves regardless of fiscal stance. The government's budget position matters less when private lifecycle savings behavior dominates external balance determination. Financial openness, rather than restoring fiscal transmission, may amplify non-fiscal channels (investment income, portfolio rebalancing) that further loosen the fiscal--CA link.

## 5.3 Demographic Tercile Heterogeneity

The tercile analysis delivers the paper's most striking result. Splitting the sample by demographic transition stage:

- **Early transition** (youngest tercile, 77 countries): fiscal coefficient $= 0.71$ ($p < 0.001$, $R^2 = 0.228$)
- **Mid transition** (78 countries): fiscal coefficient $= 0.40$ ($p < 0.001$, $R^2 = 0.137$)
- **Late transition** (oldest tercile, 55 countries): fiscal coefficient $= 0.061$ ($p = 0.073$, $R^2 = 0.091$)

The fiscal--current account link is **12 times stronger** in young economies than in aged ones. In demographically mature economies, fiscal policy has at best a marginal effect on the current account (the late-transition coefficient is only marginally significant at $p = 0.073$). This monotonic decline from 0.71 to 0.061 is the central empirical result of this paper.

## 5.4 Pension Moderation

In a 41-country subsample with pension spending data, the fiscal$\times$pension interaction is positive ($0.010$, $p = 0.083$), and the baseline fiscal coefficient turns negative ($-0.167$, $p = 0.166$). Both coefficients are only marginally significant or insignificant, so the pension moderation result should be interpreted cautiously. The directional pattern---countries with higher pension spending showing a partially restored fiscal--CA link---is consistent with pensions serving as an institutional channel through which demographics affect both fiscal and external balances, but the evidence is suggestive rather than conclusive.

# 6. Policy Implications

## 6.1 Conditioning the EBA Fiscal Coefficient

The IMF's EBA framework treats the fiscal--CA coefficient as a structural parameter, applying it uniformly across countries to assess external imbalances and prescribe policy adjustments. Our results show this coefficient varies from 0.71 to 0.061 depending on demographic stage. EBA assessments that assume a constant fiscal coefficient will:

- **Overestimate** the external adjustment achievable through fiscal consolidation in aging economies (Japan, Germany, Italy), potentially leading to excessively ambitious consolidation targets that fail to deliver the predicted CA improvement.
- **Underestimate** fiscal leverage in young economies (Sub-Saharan Africa, South Asia), missing opportunities where fiscal adjustment is most effective for external rebalancing.

A practical modification would condition the fiscal coefficient on the old-age dependency ratio or demographic transition stage, producing country-specific adjustment elasticities.

## 6.2 The Post-Dividend Fiscal Trap

Our results, combined with the CA decomposition, reveal a compounding challenge for post-dividend economies. In the pre-dividend phase, countries benefit from a demographic tailwind (a growing working-age population generates savings surpluses) and high fiscal leverage (the 0.71 coefficient means fiscal policy efficiently corrects external imbalances). In the post-dividend phase, both advantages disappear simultaneously: the demographic tailwind reverses as retirees dissave, and fiscal leverage over external balances collapses to near zero.

Paradoxically, this makes fiscal discipline *more* critical, albeit for different reasons. In young economies, fiscal consolidation efficiently improves the current account. In aging economies, fiscal consolidation is one of the few remaining macro instruments under direct policy control---not because it works well, but because the demographic channel has turned adverse and alternatives (exchange rate adjustment, macroprudential policy) face their own constraints in aging contexts. Countries in the late-transition tercile cannot rely on demographic forces to generate surpluses and cannot rely on fiscal policy to efficiently correct deficits.

## 6.3 Does Financial Openness Help?

One might expect financial openness to restore fiscal transmission by widening the channels through which fiscal adjustment reaches external balances. However, our KAOPEN probe (Table R5) finds the opposite: within aging economies, fiscal$\times$KAOPEN is *negative* ($-0.068$, $p = 0.003$), meaning openness further loosens the fiscal--CA link. Financial openness in aging economies amplifies non-fiscal channels (investment income, portfolio rebalancing) that are orthogonal to fiscal policy, effectively substituting financial-cycle dynamics for fiscal transmission. This result reinforces the fiscal leverage trap: neither demographic forces nor financial openness offer an escape.

## 6.4 Pension Reform as Suggestive Avenue

The pension moderation result ($p = 0.083$, 41-country subsample) is suggestive of one potential avenue. If pension systems that explicitly link fiscal contributions to foreign asset accumulation (funded systems with international mandates) could bridge the channel mismatch, they would address both fiscal sustainability and external balance dimensions simultaneously. However, this hypothesis requires substantially better institutional data than currently available and should be treated as a direction for future research.

# 7. Conclusion

This paper demonstrates that demographic structure dramatically moderates the twin deficits relationship. The fiscal--current account coefficient declines monotonically from 0.71 in young economies to 0.061 in aged ones---a factor of twelve. This moderation operates through a channel mismatch: the Z demographic vector transmits through the non-trade CA component ($Z_1 = 41.4$, $p = 0.007$), while fiscal policy operates through aggregate demand and the trade balance. Fiscal leverage is directly stronger when trade dominates the CA (fiscal$\times$trade-dominance = $0.262$, $p < 0.001$; Table R4). As populations age, the absolute amplitude of external balance variation collapses and fiscal policy targets a lower-amplitude trade margin in economies where non-fiscal dynamics increasingly matter.

Importantly, demographics do not confound the twin-deficit relationship (0.8% attenuation), so the fiscal--CA correlation is genuine. But the transmission coefficient is endogenous to demographic structure. The practical implication is that the IMF's EBA framework should condition its fiscal coefficient on demographic transition stage. More broadly, post-dividend economies face a fiscal leverage trap: the simultaneous reversal of the demographic tailwind and erosion of fiscal effectiveness creates a self-reinforcing challenge for external adjustment.

The pension moderation result ($p = 0.083$, 41-country subsample) is suggestive at best and should be treated as a hypothesis for future work rather than a policy conclusion. A definitive test would require institutional data on funded versus pay-as-you-go pension architecture, the share of pension assets allocated internationally, and the fiscal accounting treatment of contributions---none of which our panel provides at sufficient coverage.

# Appendix

## Table A1: Non-Trade CA Residual vs. Official BOP Income --- Sources of Divergence

| Dimension | Non-trade residual (CA $-$ Trade) | Official BOP income (WDI) |
|-----------|--------------------------------|---------------------------|
| **Definition** | Everything in CA not captured by goods & services trade | Net primary income + net secondary income |
| **Components included** | Primary income, secondary income, capital transfers, statistical discrepancy | Primary income, secondary income only |
| **Statistical discrepancy** | Absorbed into residual | Excluded |
| **Data source for CA** | IMF WEO | IMF BOP |
| **Data source for trade** | WDI (NE.RSB.GNFS.ZS) | --- |
| **Country-year coverage** | 3,907 obs, 150 countries (regression sample) | 4,086 obs, 160 countries |
| **Level correlation** | --- | $r = 0.950$ ($N = 5{,}748$) |
| **Z$_1$ coefficient** | $+41.4$ ($p = 0.007$, $R^2 = 0.129$) | $-45.4$ ($p = 0.039$, $R^2 = 0.017$) |
| **Z$_1$ on primary income** | --- | $-31.1$ ($p = 0.009$, $R^2 = 0.007$) |
| **Z$_1$ on secondary income** | --- | $+3.9$ ($p = 0.738$, $R^2 = 0.182$) |

The high level correlation ($r = 0.950$) confirms the residual is dominated by non-trade flows and is not measurement noise. The sign reversal in regression coefficients likely reflects: (i) the residual absorbing statistical discrepancy correlated with demographic structure; (ii) different effective samples due to BOP data sparsity in developing countries; and (iii) the mechanical link between the residual and CA that is absent from the independently measured BOP income series. The trade balance channel is unambiguously null for Z under both approaches.

## Table A2: Granger-Style Horizon Analysis --- Z$_1$ $\to$ Future $\Delta$KAOPEN

| Horizon | Z$_1$ Coef | p-value | R$^2$ |
|---------|---------|---------|-----|
| t+1 | 0.231 | 0.238 | 0.002 |
| t+2 | 0.179 | 0.347 | 0.005 |
| t+3 | 0.130 | 0.487 | 0.005 |
| t+4 | 0.108 | 0.560 | 0.004 |
| t+5 | 0.015 | 0.935 | 0.005 |

No significant predictability at any horizon. Near-zero $R^2$ at all horizons confirms demographics predict the *level* of capital account openness, not discrete liberalization events.

## Table R1: Winsorization Robustness --- CA Decomposition

See phase6_winsorization.md. Key result: Z$\to$trade null and OADR$\to$trade survive all winsorization levels (p1/p99, p5/p95). Z$\to$non-trade weakens with winsorization (41.4*** $\to$ 25.7* $\to$ 11.4 NS), consistent with influential creditor-economy observations.

## Table R2: CA Composition by Demographic Tercile

| Tercile | Label | N | Countries | Mean OADR | Trade Share | |Trade/GDP| | |NonTrade/GDP| |
|---------|-------|---|-----------|-----------|-------------|-----------|--------------|
| 1 | Early (young) | 2,392 | 97 | 0.1% | 0.594 | 15.07 | 11.73 |
| 2 | Mid | 2,121 | 90 | 0.1% | 0.583 | 11.94 | 8.73 |
| 3 | Late (old) | 1,993 | 60 | 0.2% | 0.590 | 7.05 | 4.37 |

Trade shares are approximately constant across terciles; the declining magnitude of *both* components in aging economies explains why the share ratio is preserved while the fiscal channel weakens in absolute terms.

## Table R3: Trade Share Regression

OADR$\to$trade share: $0.113$, $p = 0.261$ (NS). Z$_1$$\to$trade share: $-0.031$, $p = 0.912$ (NS). The trade share does not systematically decline with aging in a regression framework, though both absolute components decline sharply (Table R2).

## Table R4: Fiscal Coefficient by Trade Dominance

| Specification | Fiscal Coef | Interaction | N | R$^2$ |
|---------------|-------------|-------------|---|-------|
| Baseline | 0.557*** | --- | 3,907 | 0.250 |
| fiscal $\times$ trade-dominant (indicator) | 0.406*** | +0.262*** | 3,907 | 0.258 |
| fiscal $\times$ trade-share (continuous) | 0.288*** | +0.427*** | 3,907 | 0.255 |
| Subsample: trade-dominant | 0.310*** | --- | 2,010 | 0.350 |
| Subsample: non-trade-dominant | 0.528*** | --- | 1,897 | 0.099 |

The positive interaction confirms fiscal leverage is stronger when trade dominates the CA.

## Table R5: 3$\times$3 Factorial --- Fiscal Coefficient by Demo $\times$ KAOPEN Tercile

| Demo | KAOPEN | Fiscal Coef | p-value | N |
|------|--------|-------------|---------|---|
| Young | Closed | 0.544*** | <0.001 | 682 |
| Young | Mid | 0.359*** | <0.001 | 484 |
| Young | Open | 1.006*** | <0.001 | 410 |
| Mid | Closed | 0.297*** | 0.005 | 550 |
| Mid | Mid | 0.387*** | <0.001 | 633 |
| Mid | Open | 0.582*** | <0.001 | 243 |
| Old | Closed | 0.060 | 0.757 | 71 |
| Old | Mid | 0.172*** | 0.009 | 419 |
| Old | Open | $-0.018$ | 0.626 | 914 |

Financial openness does *not* restore fiscal leverage in aging economies. Within the late-transition tercile, fiscal$\times$KAOPEN = $-0.068$ ($p = 0.003$). Note: R$^2$ values are GLS pseudo-R$^2$ and can be negative when within-group variation is dominated by entity effects in small cells (e.g., Old $\times$ Mid, $N = 419$, 20 countries).

## Table R6--R7: KAOPEN Interaction Within Demographic Terciles

Within-tercile fiscal$\times$KAOPEN: Young $+0.085$ ($p = 0.028$); Mid $+0.043$ ($p = 0.321$); Old $-0.068$ ($p = 0.003$). The sign reversal from positive (young) to negative (old) shows that openness amplifies fiscal leverage in young economies but further erodes it in old ones.

# References

Abbas, S. A., Bouhga-Hagbe, J., Fat\'as, A., Mauro, P., & Velloso, R. C. (2011). Fiscal policy and the current account. *IMF Economic Review*, 59(4), 603--629.

Auerbach, A. J., & Kotlikoff, L. J. (1987). *Dynamic Fiscal Policy*. Cambridge University Press.

Bernheim, B. D. (1988). Budget deficits and the balance of trade. *Tax Policy and the Economy*, 2, 1--31.

Chinn, M. D., & Ito, H. (2006). What matters for financial development? Capital controls, institutions, and interactions. *Journal of Development Economics*, 81(1), 163--192.

Chinn, M. D., & Prasad, E. S. (2003). Medium-term determinants of current accounts in industrial and developing countries: An empirical exploration. *Journal of International Economics*, 59(1), 47--76.

Corsetti, G., & M\"uller, G. J. (2006). Twin deficits: Squaring theory, evidence and common sense. *Economic Policy*, 21(48), 597--638.

Gourinchas, P.-O., & Rey, H. (2007). International financial adjustment. *Journal of Political Economy*, 115(4), 665--703.

Gourinchas, P.-O., & Rey, H. (2014). External adjustment, global imbalances, valuation effects. In G. Gopinath, E. Helpman, & K. Rogoff (Eds.), *Handbook of International Economics* (Vol. 4, pp. 585--645). Elsevier.

Higgins, M. (1998). Demography, national savings, and international capital flows. *International Economic Review*, 39(2), 343--369.

IMF. (2019). *External Balance Assessment Methodology: 2018 Update*. IMF Working Paper 19/65.

Koomen, M., & Wicht, A. (2023). Demographics and current account imbalances. *Journal of International Economics*, 144, 103792.

Lane, P. R., & Milesi-Ferretti, G. M. (2007). The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970--2004. *Journal of International Economics*, 73(2), 223--250.

Lee, R., & Mason, A. (2011). *Population Aging and the Generational Economy: A Global Perspective*. Edward Elgar.

Obstfeld, M., & Rogoff, K. (2005). Global current account imbalances and exchange rate adjustments. *Brookings Papers on Economic Activity*, 2005(1), 67--146.

Phillips, S., et al. (2013). The External Balance Assessment (EBA) Methodology. *IMF Working Paper* 13/272.
